Date of Award

Spring 5-2017

Embargo Period


Degree Type


Degree Name

Doctor of Philosophy (PhD)




Onur Mutlu


Over the past two decades, the storage capacity and access bandwidth of main memory have improved tremendously, by 128x and 20x, respectively. These improvements are mainly due to the continuous technology scaling of DRAM (dynamic random-access memory), which has been used as the physical substrate for main memory. In stark contrast with capacity and bandwidth, DRAM latency has remained almost constant, reducing by only 1.3x in the same time frame. Therefore, long DRAM latency continues to be a critical performance bottleneck in modern systems. Increasing core counts, and the emergence of increasingly more data-intensive and latency-critical applications further stress the importance of providing low-latency memory accesses. In this dissertation, we identify three main problems that contribute significantly to long latency of DRAM accesses. To address these problems, we present a series of new techniques. Our new techniques significantly improve both system performance and energy efficiency. We also examine the critical relationship between supply voltage and latency in modern DRAM chips and develop new mechanisms that exploit this voltage-latency trade-o to improve energy efficiency. First, while bulk data movement is a key operation in many applications and operating systems, contemporary systems perform this movement inefficiently, by transferring data from DRAM to the processor, and then back to DRAM, across a narrow o -chip channel. The use of this narrow channel for bulk data movement results in high latency and high energy consumption. This dissertation introduces a new DRAM design, Low-cost Inter-linked SubArrays (LISA), which provides fast and energy-efficient bulk data movement across sub- arrays in a DRAM chip. We show that the LISA substrate is very powerful and versatile by demonstrating that it efficiently enables several new architectural mechanisms, including low-latency data copying, reduced DRAM access latency for frequently-accessed data, and reduced preparation latency for subsequent accesses to a DRAM bank. Second, DRAM needs to be periodically refreshed to prevent data loss due to leakage. Unfortunately, while DRAM is being refreshed, a part of it becomes unavailable to serve memory requests, which degrades system performance. To address this refresh interference problem, we propose two access-refresh parallelization techniques that enable more overlap- ping of accesses with refreshes inside DRAM, at the cost of very modest changes to the memory controllers and DRAM chips. These two techniques together achieve performance close to an idealized system that does not require refresh. Third, we find, for the first time, that there is significant latency variation in accessing different cells of a single DRAM chip due to the irregularity in the DRAM manufacturing process. As a result, some DRAM cells are inherently faster to access, while others are inherently slower. Unfortunately, existing systems do not exploit this variation and use a fixed latency value based on the slowest cell across all DRAM chips. To exploit latency variation within the DRAM chip, we experimentally characterize and understand the behavior of the variation that exists in real commodity DRAM chips. Based on our characterization, we propose Flexible-LatencY DRAM (FLY-DRAM), a mechanism to reduce DRAM latency by categorizing the DRAM cells into fast and slow regions, and accessing the fast regions with a reduced latency, thereby improving system performance significantly. Our extensive experimental characterization and analysis of latency variation in DRAM chips can also enable development of other new techniques to improve performance or reliability. Fourth, this dissertation, for the first time, develops an understanding of the latency behavior due to another important factor { supply voltage, which significantly impacts DRAM performance, energy consumption, and reliability. We take an experimental approach to understanding and exploiting the behavior of modern DRAM chips under different supply voltage values. Our detailed characterization of real commodity DRAM chips demonstrates that memory access latency reduces with increasing supply voltage. Based on our characterization, we propose Voltron, a new mechanism that improves system energy efficiency by dynamically adjusting the DRAM supply voltage based on a performance model. Our extensive experimental data on the relationship between DRAM supply voltage, latency, and reliability can further enable developments of other new mechanisms that improve latency, energy efficiency, or reliability. The key conclusion of this dissertation is that augmenting DRAM architecture with simple and low-cost features, and developing a better understanding of manufactured DRAM chips together leads to significant memory latency reduction as well as energy efficiency improvement. We hope and believe that the proposed architectural techniques and detailed experimental data on real commodity DRAM chips presented in this dissertation will enable developments of other new mechanisms to improve the performance, energy efficiency, or reliability of future memory systems.